CN110148133A - Circuit board relic image-recognizing method based on characteristic point and its structural relation - Google Patents
Circuit board relic image-recognizing method based on characteristic point and its structural relation Download PDFInfo
- Publication number
- CN110148133A CN110148133A CN201910534181.9A CN201910534181A CN110148133A CN 110148133 A CN110148133 A CN 110148133A CN 201910534181 A CN201910534181 A CN 201910534181A CN 110148133 A CN110148133 A CN 110148133A
- Authority
- CN
- China
- Prior art keywords
- image
- point
- matching
- feature points
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 82
- 239000012634 fragment Substances 0.000 claims description 79
- 238000009826 distribution Methods 0.000 claims description 14
- 238000003860 storage Methods 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 4
- 239000003086 colorant Substances 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000003491 array Methods 0.000 claims 1
- 230000001174 ascending effect Effects 0.000 abstract 1
- 230000000717 retained effect Effects 0.000 abstract 1
- 230000008569 process Effects 0.000 description 13
- 239000002360 explosive Substances 0.000 description 11
- 238000010586 diagram Methods 0.000 description 8
- 238000004880 explosion Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 230000000875 corresponding effect Effects 0.000 description 5
- 238000000605 extraction Methods 0.000 description 5
- 238000005070 sampling Methods 0.000 description 5
- 238000003709 image segmentation Methods 0.000 description 4
- 230000011218 segmentation Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000003909 pattern recognition Methods 0.000 description 3
- 238000003466 welding Methods 0.000 description 3
- 101150041570 TOP1 gene Proteins 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 210000000887 face Anatomy 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 229910000679 solder Inorganic materials 0.000 description 2
- ATJFFYVFTNAWJD-UHFFFAOYSA-N Tin Chemical compound [Sn] ATJFFYVFTNAWJD-UHFFFAOYSA-N 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 229910052751 metal Inorganic materials 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000000452 restraining effect Effects 0.000 description 1
- 238000005476 soldering Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30141—Printed circuit board [PCB]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
This disclosure relates to the circuit board relic image-recognizing method based on characteristic point and its structural relation, wherein preserving sample image and its characteristic point and feature description in sample database, the method comprising the steps of: 1, extracting the characteristic point of relic image；2, the SIFT description of characteristic point is calculated；3, the first Euclidean distance of the SIFT description of each characteristic point and the SIFT description of each characteristic point of each sample image is calculated separately；4, to the ascending sequence of the first Euclidean distance, the first Euclidean distance is selected in feature point set to sort forward multiple characteristic points, as this feature point candidate matches point；5, the second Euclidean distance between candidate matches point and the Shape context description of this feature point is calculated, the candidate matches point that the second Euclidean distance is not more than predetermined threshold, the matching point set as this feature point are retained；6, matched second hub-and-spoke configuration of the first hub-and-spoke configuration with relic image is found in sample image；7, according to the similitude between hub-and-spoke configuration, the matching degree of sample image and relic image is obtained.
Description
Technical Field
The present disclosure relates to the field of computer image processing technologies, and more particularly, to a method for identifying circuit board fragments of an explosive device.
Background
The principle of the image recognition of the circuit board fragments is as follows: after the circuit board residue image is acquired (for example, after an explosive device (after explosion) is acquired by a public security agency, the circuit board residue image is shot as an image), the circuit board residue image belongs to which model of circuit board original plate image in the circuit board original plate image library, and therefore, the model of the circuit board from which the circuit board residue is originated can be determined.
The identification of the image of the broken pieces of the circuit board mainly relates to two important problems of the image segmentation of the circuit board and the identification of the image of the broken pieces: 1) circuit board image segmentation (region-of-interest extraction), namely, automatically extracting regions of interest where components, characters, welding spots and the like are located from the circuit board image, and is the basis of circuit board image description and residue image identification; 2) the image recognition of the circuit board fragments belongs to a practical application of image pattern recognition.
In the field of image recognition, researchers have achieved many achievements, such as recognition of image objects such as faces, fingerprints, characters, license plates, and the like, and have been applied to actual life; but it is not uncommon to investigate the identification of circuit board residue by image recognition techniques. Different from a common image object identification task, the circuit board fragments are only small parts of the whole circuit board, and the fragment images do not have all the characteristics of the whole image object; in the identification process, the whole body can be identified only according to limited local description, which belongs to few special research problems in the field of image pattern identification, so that the research related to the invention has certain originality.
The following illustrates the prior art solution in order to understand the background of the present invention.
Technical scheme of prior art I
In the field of traditional image pattern recognition, image recognition by using local feature points is a good strategy, and some stably existing feature points can represent local information of an image. The matching method based on the local characteristic points comprises the steps of respectively extracting the object to be identified and the local characteristic point description in the sample set image, and identifying by using a proper matching strategy. The local feature point comparison method proposed by Sushkov [1] et al is called BPM (binary matching) method, and the similarity of SIFT feature point [2] feature description and the triangular structure constraint formed by similar feature points are utilized to carry out image comparison and identification; for the general image recognition problem, the method has high recognition accuracy.
Disadvantages of the first prior art
Li think [3] improves the matching method based on local feature points proposed in the document [1] and is used for identifying the image of the circuit board residue, and can identify the residue image with more local features more accurately; the following problems still remain:
(1) SIFT feature points extracted from the circuit board image are not all stable key points, and unstable feature points may be extracted from the texture of the circuit board bottom plate, the metal of components and the like due to illumination and shooting angles;
(2) in the feature point description matching process, only the similarity of SIFT description is considered, and the strict structural relationship among feature points is not considered, so that mismatching of the feature points can be caused;
(3) the constraint of similarity of triangles is weak, and the structural relationship between matched triangles is not considered, so that mismatching between different circuit boards with similar structures can be caused;
(4) if stable feature points do not exist in the circuit board image, for example, some back panels only have soldering tin with different shapes, the matching method based on the SIFT description of the feature points is difficult to realize accurate matching of the fragment objects.
Disclosure of Invention
In view of the above problems in the prior art, the inventor proposes the present invention, which relates to an automatic identification method for circuit board residue based on feature points and component structure relationship, and is capable of accurately identifying different types of residue images. The identification method is applied to tracing the source of the circuit board fragments of the explosion device in the explosion case, so that a large amount of time and labor required by manual arrangement comparison can be saved, and important technical support is provided for the work of detecting and breaking the explosion case by a public security organization.
According to the embodiment of the invention, the circuit board fragment image identification method based on the characteristic points and the structural relationship thereof is provided, wherein the sample image I corresponding to the known circuit board is stored in the sample libraryiAnd characteristic points and characteristic description thereof, wherein the fragments of the circuit board of the explosive device correspond to a fragment image I0The method comprises the following steps:
step 1, extracting a fragment image I0The characteristic points of (1);
step 2, calculating SIFT description of the feature points to form a feature point set S0；
Step 3, respectively calculating the characteristic point set as S0With the SIFT description of each feature point in each sample image IiCharacteristic point set S ofiThe first euclidean distance of the SIFT description of each feature point in (a);
step 4, sorting the first Euclidean distances from small to large, and selecting a feature point set SiThe first Euclidean distance of the plurality of characteristic points in the sequence of S is taken as0A candidate matching point of the feature point of (1);
step 5, calculating candidate matching points and S0A second Euclidean distance between the shape context descriptions of the feature points, and candidate matching points with the second Euclidean distance not greater than a predetermined threshold are reserved as S0The feature ofA set of matching points of points;
step 6, establishing a fragment image I0And in the sample image IiSearching a second star structure matched with the first star structure, wherein the second star structure is established based on the matching point set;
step 7, obtaining each sample image I according to the similarity of the second star structure and the first star structureiAnd a broken picture image I0The degree of matching of (2).
The invention can be used in tracing the source of the fragments of the explosive device and provides technical support for detecting the explosive cases. The image processing and pattern recognition technology is used for identifying the circuit board fragments of the explosion device, original boards of the same type are searched out, a large amount of time and energy required for manual groping by professionals are saved while the identification accuracy is guaranteed, and the method has important significance for fighting against explosion crimes.
Drawings
FIG. 1 is a schematic diagram of a method for identifying a circuit board residue image based on feature points and their structural relationships according to an embodiment of the present invention;
FIGS. 2 and 3 are diagrams illustrating the concept of class shape context according to embodiments of the present invention;
fig. 4 to 6 are schematic flow charts of a method for identifying a circuit board residue image based on feature points and their structural relationships according to an embodiment of the present invention;
FIGS. 7 and 8 are conceptual illustrations of star configuration matching according to embodiments of the present invention;
FIG. 9 is a functional block diagram of an explosive device circuit board debris identification system according to an embodiment of the present invention;
fig. 10 is a schematic diagram of an operating environment of a system in which an application is installed according to an embodiment of the present invention.
Detailed Description
The following describes the embodiments in further detail with reference to the accompanying drawings.
It will be appreciated by those skilled in the art that while the following description refers to numerous technical details of embodiments of the present invention, this is by way of example only, and not by way of limitation, to illustrate the principles of the invention. The present invention can be applied to places other than the technical details exemplified below as long as they do not depart from the principle and spirit of the present invention.
In addition, in order to avoid limiting the description of the present specification to a great extent, in the description of the present specification, it is possible to omit, simplify, and modify some technical details that may be obtained in the prior art, as would be understood by those skilled in the art, and this does not affect the sufficiency of disclosure of the present specification.
Hereinafter, an embodiment for carrying out the present invention will be described. Note that the description will be given in the following order: 1. summary of the inventive concept (fig. 1); 2. a circuit board fragment image identification method (fig. 2 to 8) based on the feature points and the structural relationship thereof; 3. an explosive device circuit board debris identification system (fig. 9); 4. a system with an application installed according to an embodiment of the present invention (fig. 10).
1. Summary of the invention
The method for identifying the circuit board fragments designed and realized by the invention simultaneously considers the similarity of feature description and the consistency of the space structure of the feature points, the core idea of the identification algorithm is to construct and search a star-shaped structure matching pair, and document [3] adopts a parallelogram structure based on the feature point pair to search a matching area, so that the method has not strong constraint on the space structure of the feature points and is easy to generate the condition of mismatching. In order to find a matching area more accurately, the invention adopts a comparison method based on a star structure, and the matching pair of the star structure needs to meet the description consistency, the space consistency and the structure consistency; meanwhile, in order to eliminate mismatching among similar structures on different types of circuit boards, the high-level matching criterion is adopted as follows: the spatial structure between the matching sets of star-like structures is also consistent and can be considered as matching.
The comparison process based on the two comparison methods of the feature point structural relationship and the component structural relationship is basically similar, and the difference lies in that (1) in the comparison process of the feature point structure is used as a node to construct a matched star-shaped structure, and in the comparison process of the component structure, each component in a circuit board image segmentation result is used as a node to construct a matched structure; (2) the former node features SIFT description and class shape context, and the latter features size, shape, color, texture and other appearance information of the component.
Because the residue of the circuit board is far smaller than the original board, the available image area in the identification is very limited, and in order to obtain an accurate residue identification result, the invention fully considers the characteristics of each level of the residue when designing the residue identification scheme, provides a scheme for carrying out residue comparison and identification in a hierarchical manner, and carries out residue comparison and identification in a hierarchical manner by using a strategy of combining local characteristics with global characteristics.
The comparison process is divided into the following three layers: comparing the first layer, and inspecting the similarity of local feature points based on feature description; the second layer of comparison, which is to introduce a star structure formed by the characteristic points for comparison by utilizing the rule that the structures of the characteristic points of the matched images must be the same, and identify the matched circuit board images according to the fragments; and the third layer of comparison belongs to a high-layer matching strategy, the identification problem is considered from the overall situation of the fragment image, and the constraint on the matching image is as follows: the spatial distribution structure among all the matched star structures on the original plate image of the matching circuit is consistent with the spatial distribution structure among all the star structures finding the matching pairs on the residual image, and the matching accuracy is ensured from the perspective of higher level and overall situation.
The circuit board fragments have diversity, and a plurality of stable characteristic points exist in some fragment images, such as front images of the circuit board; and some residual images have few or no stable characteristic points, such as the back images of the circuit boards, sometimes only welding points exist, and the stable characteristic points are difficult to extract. Therefore, the invention adopts a comparison identification method based on the combination of the characteristic points and the structural relationship thereof for the circuit board image which can extract enough stable characteristic points; for the circuit board image lacking stable characteristic points, a comparison identification method based on combination of components and structural relations thereof is adopted; therefore, different fragments can use a proper comparison method to achieve the purpose of identifying the matched overall image according to the local image.
Specifically, for a circuit board image capable of extracting stable feature points, Harris angular points [4] of the image are extracted as compared feature points, SIFT description and class shape context description of each feature point are calculated and used as local feature description; and comparing local features according to the local features to identify similar feature points; and constructing a star structure by taking the characteristic points as nodes, and searching a matched star structure among the compared images. For the fragment image without stable feature points, a fragment identification method based on a component structure relation is adopted, a component segmentation result of the fragment image is obtained by utilizing a circuit board image segmentation algorithm, the size, the shape and the color of the component are used as feature descriptions of the component, and local feature comparison is carried out according to the feature descriptions so as to identify similar components; and constructing a star structure by taking the components as nodes, and searching a matched star structure between the compared images.
The same components can exist on different circuit boards, and the same components can generate local feature point matching and even matching of individual star-shaped structures, thereby causing mismatching of the fragments and the non-homologous circuit boards. In order to avoid mismatching between different kinds of circuit boards with similar structures, a high-level matching strategy requires that consistency of space distribution structures must be met among all matched star structures.
The invention fully considers the description of the characteristic points and the topological structure relationship among the characteristic points, wherein, the advantage of using the star-shaped structure matching constraint of the characteristic points is as follows: (1) one star structure can be associated with a plurality of characteristic points, and only 3 characteristic points are associated unlike a triangular structure, so that stronger space structure constraint is provided for matching characteristic points; (2) more flexible than a triangular structure. The method can accurately identify different types of circuit board fragments, and can carry out the test of automatic fragment image identification in a data set containing 2707 effective fragments, wherein the accuracy rate of Top1 of the method reaches 91.9 percent, and the accuracy rate of Top5 of the method reaches 96.0 percent.
Fig. 1 is a general flow diagram of a circuit board residue image identification method based on feature points and their structural relationships according to an embodiment of the present invention.
As shown in fig. 1, the method for identifying the image of the circuit board residue based on the feature points and the structural relationship thereof provided by the invention can be divided into the following two stages:
1) in the stage of establishing the sample library, component segmentation and feature point extraction operations need to be performed on each image in the sample library, and finally feature descriptions are calculated and stored in the sample library.
2) In the stage of comparing and identifying the fragments, firstly, judging whether stable characteristic points exist in the fragment image, if a plurality of stable characteristic points exist in the fragment, extracting the characteristic points of the fragment image, and otherwise, carrying out component segmentation on the fragment image; and then, calculating the feature description of the feature points or the components, and finally comparing the feature description with each sample library image in the sample library to judge whether the images are matched with each other or not.
SIFT feature points are adopted in comparison methods based on feature points in documents [1] and [3], but a large number of unstable key points exist in the SIFT feature points extracted from a circuit board image, and the accuracy and the operation speed of a comparison algorithm are greatly influenced. The Harris angular point [4] of the image can well represent the angular point in the image, has the advantages of insensitivity to brightness and contrast change, rotation invariance and the like, and is more suitable for being used as a characteristic point in circuit board image recognition, so the Harris angular point is used as the characteristic point of the circuit board image.
In the Harris angular point extraction process, because the areas presented by the angular points at different positions on the circuit board are different in size, the angular point of a character with a fine stroke can be determined in a small area, the angular point of a large component can be determined in a large area, and the traditional Harris angular point detection algorithm can only detect characteristic points under a single scale and does not have scale invariance. Therefore, the invention adopts a multi-scale Harris angular point detection method, integrates angular point results detected under different scales, determines information such as position, response, main direction and the like of the Harris angular point, and has scale invariance. The angular point detection algorithm is applied to a circuit board image, and the extracted Harris angular point is accurate in position, comprehensive in coverage and in line with actual characteristic point distribution of the circuit board image.
In the aspect of feature description, the invention adopts SIFT description and class shape context description as two modes:
1) SIFT description [2]
In the comparison algorithm based on the feature point structure relationship, the feature points need to be accurately and discriminatively described so as to match similar feature points and distinguish different feature points. SIFT description is insensitive to rotation, brightness, scale and the like, and has strong anti-noise capability, so that SIFT is adopted to describe local information of feature points. Extracting feature points of the fragment image to obtain information such as positions, main directions and the like of the feature points; and taking the W multiplied by W neighborhood window by taking the characteristic point as a center, solving the gradient amplitude and direction of each pixel point, and weighting the gradient amplitude of each pixel point in the window by utilizing a Gaussian function. The pixels in the window are divided into 16 blocks, 8 directional gradient distributions are counted on each small block, namely the weighted gradient amplitude values are accumulated in all directions, the counted values are cascaded to form a 128-dimensional characteristic vector, normalization processing is carried out, and the influence of illumination is removed. SIFT description accurately reflects local information of the feature points, and is an important basis for the comparison stage of the feature points
2) Class shape context description
The most prominent difficulty of the image recognition of the circuit board fragments is as follows: the area of the residual is small, the characteristic information is less, and the similar relation of local description of a single characteristic point is only used, so that the error matching of different characteristic points is easily caused; in order to be able to accurately identify the debris image, in addition to using the description information of the local feature points, the spatial distribution structural relationship of the circuit board feature points must be used to constrain the matching of the feature points so as to suppress the mismatching of the feature points as much as possible. Considering the homologous circuit board image, the spatial distribution of the characteristic points is consistent; for this purpose, shape context characteristics are introduced into a task of identifying the image of the circuit board fragment so as to describe the spatial distribution structure relationship of the characteristic points.
The shape context feature [5] describes the contour structure information of the object, encodes the relative positions of points on the contour of the object, and extracts the structure information between contour points to describe the overall structure of the object. The extraction method of the shape context description is shown in fig. 2, in which (a) is a character image; (b) extracting a schematic diagram for the shape context, wherein black points in the diagram are a group of discrete point sets obtained by sampling the edges of the character outline at equal intervals, N concentric circles are established at equal intervals according to logarithmic distance by taking one point as a center, and the region is equally divided along the circumference M; and (c) counting the number of edge points falling in each region to obtain a contour sampling point histogram under a polar coordinate system, wherein the color of each small block represents the number of edge sampling points falling in the region, and the deeper the color is, the more the number of edge points contained in the region is. The two-dimensional histogram is the shape context description of the contour sampling points positioned at the circle center position in the step (b), and the shape context description characteristics are calculated for all the sampling points of the character contour, so that the shape context characteristics for describing the character contour structure are obtained.
In order to describe the spatial distribution structure relationship of the image characteristic points of the circuit board and realize quick comparison and identification, the invention improves the original shape context description characteristics, and replaces a rectangular coordinate system with a logarithmic polar coordinate system to count the spatial distribution of each characteristic point in a certain window size taking each characteristic point as the center so as to describe the structure relationship between adjacent characteristic points. The method comprises the following specific steps: as shown in fig. 3, a 49 × 49 neighborhood window is taken with the current feature point as the center, the pixels in the window are equally divided into 7 × 7 blocks, the size of each small block is 7 × 7, the number of feature points in each small block is counted, and finally the class shape context description of the 49-dimensional feature point is obtained, the feature describes the spatial structure relationship between the feature point and other feature points adjacent to the feature point, so that the feature is named as the class shape context feature; when the feature points are matched, the class shape context features of the matched feature point pairs are required to be similar enough, the matched feature points are constrained from the structural relationship of the spatial distribution of the feature points, and the error matching caused by the similarity of SIFT descriptions of single feature points is restrained.
As described below, the present invention also utilizes star-structured matching as a high-level matching strategy.
In the stage of matching the feature points, the similarity of the individual feature points is considered, and accordingly, the accuracy of the matching point pairs cannot be completely guaranteed. Because, even for the same model of circuit board, different physical object fabrication processes may result in non-hundred percent similarity between them; the circuit boards of different models have the same component area, which causes the circuit boards of different models to have similar areas; in addition, various interference factors exist during image acquisition; therefore, it is not sufficient to rely only on the similarity of the respective feature points; it was also found in the experiments that there were cases where the similarity score between unmatched feature points was higher than the similarity score between the true matched feature points.
In the real world, people need to recognize faces and observe the whole and details of the face region, and can not only rely on one eye or nose to judge that the person is a certain person; this also shows that, when an object is identified by an image, it is necessary to examine not only the similarity of key parts but also whether the relationship existing between the similar parts is matched. Based on the method, in the process of designing the fragment image identification method, the structural relationship among the feature points is used for restraining the matched feature points so that the feature points can be correctly matched and wrong matched pairs are eliminated; in the invention, the star is used in consideration of the diversity of the appearance positions of the image characteristic points of the circuit boardThe type structure describes the topological structure relationship among the characteristic points, and the characteristic points which are matched with each other need to satisfy the similarity constraint of local description characteristics and the similarity constraint of a star structure relationship formed by other characteristic points. Firstly, selecting feature points positioned in a non-edge area of a fragment as a central node to construct a star-shaped structure in a feature point set of the fragment image; secondly, for any sample image Ii(i ═ 1, …, N), in its feature point set SiIf a plurality of star structure matching pairs exist, the matching pairs also meet the constraint of space structure consistency, and the condition is called as a high-level matching criterion.
The following describes an implementation of the method for identifying a circuit board residue image based on feature points and their structural relationships in the present invention.
2. Circuit board fragment image identification method based on feature points and structural relationship thereof
As described above, the method for identifying the image of the circuit board residue based on the feature points and the structural relationship thereof according to the present invention includes a sample library establishing stage and a residue comparison and identification stage, i.e., before the residue comparison and identification, the sample library is required to be established.
1) In the stage of establishing a sample library, the method for identifying the circuit board fragment image based on the feature points and the structural relationship thereof comprises the following steps:
s100, extracting and calculating each sample image IiForming SIFT description and class shape context description of the feature points;
s200, storing the characteristic description into a sample library (bottom library), namely, storing each sample image I in the bottom libraryiSIFT descriptions, each with corresponding feature points, and class shape context descriptions.
Specifically, through the above steps, the feature point set recorded in the base library is Si，i＝1,…,N, wherein N is the number of circuit board original boards of different models in the sample library, SiFor the ith sample image IiThe feature point set of (1).
2) In the stage of comparing and identifying the broken pieces, the method for identifying the circuit board broken piece image based on the characteristic points and the structural relationship thereof comprises two matching stages of characteristic point matching and star-shaped structure matching,
2-1) the feature point matching stage comprises the following steps:
step S300, extracting a circuit board fragment image I to be identified0The Harris characteristic points of (1) are used as characteristic points of the circuit board image;
s400, calculating SIFT description of the Harris feature points, and forming a feature point set S0；
Step S500, respectively calculating a feature point set as S0With the SIFT description of each feature point in each sample image IiCharacteristic point set S ofiThe euclidean distance of the SIFT description of each feature point in (a);
step S600, for the fragment image I0Characteristic point set S of0The following steps are performed for each feature point of (1):
step S610, traversing sample image IiCharacteristic point set S ofiCalculating S0The characteristic point of (1) and SiThe Euclidean distance of SIFT description of each feature point in S is selectediA plurality of feature points (e.g., 15) having the smallest euclidean distance therebetween as S0A candidate matching point of the feature point of (1);
in this way, the similarity of local description between matching point pairs can be ensured;
step S620, calculating candidate matching points and S0Retaining candidate matching points having a euclidean distance of not more than a predetermined threshold (e.g., 5) as S0The set of matching points of the feature points;
the step is used for constraining the matching feature points to be consistent in spatial distribution of the feature points in the neighborhood, and the remaining feature points form a candidate matching point set of the feature points of the debris. The characteristic point matching comprehensively considers SIFT description and class shape context description of the characteristic points, and determines matching point pairs according to local description similarity and characteristic point space distribution structure relationship similarity of the characteristic points so as to ensure accuracy of the method.
By being S0Repeating the above steps S610 and S620 for each feature point to obtain S0Of each feature point of (1), wherein S0The characteristic point in (1) is denoted as pk(K is 1, … K, K is the number of feature points of the debris image).
Thus, for the remnant image I0Each feature point of (1), each sample image IiEach resulting in no more than 15 best matched feature points (some sample images are discarded because of insufficient matching points).
2-2) entering the star structure matching stage.
2-2-1) first, a fragment image I is created0The star-shaped structure of the characteristic points of (1).
Specifically, the method comprises the following steps:
step S700, initializing the current feature point pkLet k equal to 1;
step S800, if the current characteristic point pkIf the image is not in the edge area of the image and does not belong to the determined effective star structure, the image is taken as the center point of the star structure, otherwise, the next characteristic point p is takenk+1As the current feature point, repeating the step;
step S900, adding S0And the current feature point p inkIs not less than the star-shaped structure threshold Td(e.g., 5) as a characteristic point represented by pkEnd node p of central star structuree；
Step S1000, if p is usedkIs composed ofIf the number of the end nodes of the heart star structure is not less than 3, p is usedkThe central star structure is used as an effective star structure, and the current central point p is taken askIs marked asEnd node peIs marked as
Step S1100, the next characteristic point p is setk+1Returning to step S800 as the current feature point, if there is no next feature point pk+1Then all valid star structures are recorded, for example, M, where M is 1, …, M.
2-2-2) next, the center matching point of the star structure is found in the sample image.
Specifically, the method further comprises:
step S1200, in the sample image IiIs obtained (i.e., obtained after step S620 corresponding to S)0，SiSubset of (d), traverseThe matching point set (M) takes each feature point of the matching point set, which is not at the edge, as the central matching point (M) of the star structurec)。
2-2-3) next, finding the end matching points of the star structure in the sample image.
Different from the searching mode of the central matching point, the matching star structure has similar space structure, so that the searching of the end matching point does not need to traverse the matching point set, and only needs to be performed at peSearching in the neighborhood corresponding to the position of the sample image.
Specifically, the method further comprises:
step S1300, let pcpeIs a star-shaped structure in the image of the remnant from a central point pcTo the end point peIn the sample image IiUpper center matching point (M)c) As a starting point, establish a sum with pcpeA line segment with the same length and direction is connected, and a rectangular neighborhood (for example, 15 × 15) is taken by taking the end point of the line segment as the center, as shown in fig. 7;
step S1400, traversing all the feature points M in the rectangular neighborhoodeIf the following conditions are satisfied, the point is an end matching point:
a)Meand peThe euclidean distance described by SIFT of (a) is not more than 0.6;
b) line segment pcpeAnd line segment McMeThe included angle is less than 10 degrees, and the side length ratio of the short side to the long side is more than 0.97.
Step S1500, if a plurality of end points (for example, not less than 3) of a certain star structure of the fragment image find matching points in the sample image, a group of star structure matching candidate pairs can be formed;
in order to avoid mismatching, the similarity between the spatial structure and the regional image structure of the two star structures needs to be examined, as described below.
Step S1600, taking the star structure matching candidate pair meeting the following 4 conditions as a final star structure matching pair:
(1) the number of the successfully matched terminal points is not less than 3;
(2) the angle difference between any two groups of edges of the two matched star structures is less than 12 degrees, and the adjacent relations of the characteristic points are the same;
(3) any two groups of end points of the two matching star structures are connected, and the distance ratio of the short side to the long side is more than 0.87;
this ensures that when the side length of the star structure is long, the shift of the terminal point is not large;
(4) the Structural Similarity (SSIM) of two circumscribed rectangular image regions containing two star-shaped structures is greater than 0.3.
The first three constraints limit the space consistency (circumscribed rectangle) of the star-shaped structure, and in order to ensure the accurate matching of the star-shaped structure, the structural similarity of the matched star-shaped structure image sub-regions is further investigated, so that the SSIM characteristics are introduced for identifying similar regions. SSIM [6] measures the similarity of images from three aspects of brightness (l), contrast (c) and structure(s), and is more consistent with the visual perception of human eyes, and SSIM is the product of brightness, contrast and structure similarity:
SSIM(X,Y)＝l(X,Y)c(X,Y)s(X,Y) (1)
wherein,
in the formula, X, Y denotes images (i.e., sample image and remnant image), μXAnd muYRepresenting the mean, σ, of the imageXAnd σYRepresenting the variance, σ, of the imageXYRepresenting the covariance of images X and Y, C1, C2, C3 are preset constants.
In order to accurately calculate the SSIM of the two matching star-shaped regions, the central point of the star-shaped structure is taken as an alignment position, the equal-size external rectangle of the two star-shaped structures is taken, and the SSIM of the aligned two rectangular region images is calculated, so that the structural similarity of the region images of the two star-shaped structures is identified. The larger the SSIM value is, the higher the similarity of the two star structures is; the accuracy of the fragment comparison algorithm is greatly improved by adding SSIM limitation to the star-structure matched region.
Similar structures may exist in different kinds of circuit boards, and for such locally similar cases, the conventional feature point-based matching method may form a mismatch due to such locally similar structures.
To solve this problem, a higher-level matching criterion can be further introduced, requiring: the spatial position relationship among the matching star structure groups on the debris, the matching star structure groups with the same spatial position relationship must also exist in the matching bottom library images, as shown in fig. 8, 3 groups of star structures A, B, C exist on the debris, the center points of the star structures form a geometric shape ABC, the matching star structures on the bottom library image 1 matched with the debris are respectively a ', B ' and C ', then the high-level matching criterion is that the geometric shape a ' B ' C ' formed by the centers of the 3 star structures is the same as ABC, if the geometric shape a ' B ' C ' is satisfied, the debris image in fig. 8 is successfully matched with the bottom library image 1; although part of the star-shaped structure of the bottom library image 2 is matched with the fragment image, the geometric structure formed by the matched star-shaped structure center does not meet the high-level matching criterion, and the fragment image is not matched with the bottom library image 2. Therefore, even if the local star-shaped structures of different types of circuit boards are successfully matched with a certain star-shaped structure on the broken chip, the consistency of the spatial position relationship of a plurality of matched star-shaped structures in the high-level matching criterion is difficult to meet simultaneously, and the occurrence of mismatching conditions is greatly reduced.
2-3) sorting of matching results
Next, since one circuit board fragment image may be successfully matched with a plurality of sample library circuit board images, the matching results need to be sorted according to the characteristics of the star-shaped structure matching pairs. A fragment image I0And a sample image IiThe matching score is as follows:
wherein i is the serial number of the star-shaped structure matching pair, eiIn order to match the number of sides of the star structure,for matching the square of SIFT description distance between feature points, SSIMmax is a remnant star-shaped structure region and a sample image IiThe maximum value of structural similarity of the relevant regions. Matching scores with edge numbers and feature descriptions contained in star structureThe similarity of the images and the SSIM of the star-shaped structure area are positively correlated, the higher the matching score is, the higher the matching degree of the fragments and the images of the sample library is, and the higher the ranking is.
Therefore, the sample image with the highest matching score S can be selected, and the circuit board corresponding to the sample image is identified as the circuit board to which the explosive circuit fragments belong.
Optionally, before the step of comparing and identifying the fragments, a preprocessing step may be further included, wherein a comparison and identification scheme is selected according to the characteristics of the circuit board fragments, that is, the fragment feature points are extracted or the device segmentation operation is performed based on the feature point structural relationship or the device structural relationship, and the feature description is calculated and stored. And subsequently, describing the fragment identification process by a comparison identification scheme based on the structural relationship of the feature points, wherein the comparison process based on the structural relationship of the components is similar to the comparison process.
Specifically, for a circuit board image that cannot extract stable feature points, for example, a back board image with only solder joints, regions of component objects such as solder joints may be extracted (any known method may be used), the regions are abstracted into feature points, and features such as related colors are extracted to replace the feature points, and then the comparison process is used to identify the residue image.
In summary, the invention provides a method for identifying the residue image suitable for different circuit boards by fully considering different characteristics of the residue, and the method uses a strategy of combining the local characteristic and the global characteristic of the residue image to compare and identify the residue in a hierarchical manner. In the comparison process, the first layer of comparison is to examine the similarity of local feature points; the second layer of comparison fully utilizes the elements of which the image structures are necessarily similar, and introduces a star structure for comparison; the third layer of comparison belongs to high-level constraint, the space relationship among star structures is used as constraint, the matching accuracy is ensured, and global information is utilized. The matching method based on the feature points is suitable for fragment identification with a plurality of stable feature points, wherein the extracted feature points adopt multi-scale Harris corner points, are insensitive to brightness and contrast, have rotation and scale invariance, and accurately describe the most stable key points in circuit board identification; when feature points are subjected to feature description, SIFT is adopted to describe the similarity of local feature points, the topological structure among class shape context feature description feature points is increased, and mismatching among the feature points is effectively inhibited. And for the fragments without stable feature points, adopting a matching method based on the structural relationship of the components, and taking the components as the feature points to perform subsequent matching to supplement and perfect matching based on the feature points. In the matching stage, matching pairs with similar structural relations are established and searched based on a star structure, and the star structure matching pairs are required to meet description consistency and spatial structure consistency; and the structural similarity of the two matched image areas is restrained by using the SSIM, so that the occurrence of mismatching conditions is greatly reduced. Because similar structures may exist in different types of circuit boards, in order to avoid mismatching among the similar structures, the limitation that the space structures among the star structures are kept consistent is also increased, and the mismatching of the similar structures is effectively avoided. The overall matching method is accurate, the matching accuracy of Top1 is 91.9%, and the accuracy of Top5 reaches 96.0%.
3. Explosive device circuit board fragment identification system
Fig. 9 is a functional block diagram of an explosive device circuit board debris identification system according to an embodiment of the present invention.
The embodiment of the invention provides an explosive device circuit board fragment identification system which is used for executing the circuit board fragment image identification method based on feature points and the structural relationship thereof.
Wherein the sample library establishing module is used for extracting and calculating each sample image IiCalculating SIFT description and class shape context description of the feature points, and storing the feature description into a sample library.
The preprocessing module is used for judging whether stable characteristic points are extracted from the fragment image, if not, extracting areas of component objects such as welding spots and the like, abstracting the areas to be characteristic points, and extracting characteristic replacement characteristic points such as related colors and the like.
The characteristic extraction module is used for extracting a circuit board fragment image I to be identified0The Harris feature points are used as feature points of a circuit board image, SIFT description of the Harris feature points is calculated, and a feature point set is formed to be S0。
The characteristic comparison module is used for comparing the similarity between the characteristic points to form a matching point set.
The structure comparison module is used for further comparing star structures formed by the matching point sets, taking the spatial relationship among the star structures as constraint and outputting the final matching score of each sample image.
Furthermore, various embodiments of the invention may also be implemented as software modules or computer-readable instructions stored on one or more computer-readable media, which when executed by a processor or device component, perform the various embodiments of the invention described herein. Similarly, any combination of software modules, computer-readable media, and hardware components is contemplated by the present invention. The software modules may be stored on any type of computer-readable storage medium, such as RAM, EPROM, EEPROM, flash memory, registers, hard disk, CD-ROM, DVD, and the like.
4. System with application installed according to embodiments of the present invention
Fig. 10 illustrates an operating environment of a system in which an application is installed according to an embodiment of the present invention.
In this embodiment, the system for installing the application is installed and operated in the electronic device. The electronic device can be a desktop computer, a notebook, a palm computer, a server and other computing equipment. The electronic device may include, but is not limited to, a memory, a processor, and a display. The figure shows only an electronic device having the components described above, but it should be understood that not all of the shown components are required to be implemented, and more or fewer components may be implemented instead.
The memory may in some embodiments be an internal storage unit of the electronic device, such as a hard disk or a memory of the electronic device. The memory may also be an external storage device of the electronic apparatus in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like, provided on the electronic apparatus. Further, the memory may also include both an internal storage unit and an external storage device of the electronic apparatus. The memory is used for storing application software installed in the electronic device and various types of data, such as program codes of the system for installing the application programs. The memory may also be used to temporarily store data that has been output or is to be output.
The processor may be, in some embodiments, a Central Processing Unit (CPU), a microprocessor, or other data Processing chip, for running program code stored in the memory or Processing data, such as executing the system for installing applications.
The display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch panel, or the like in some embodiments. The display is used for displaying information processed in the electronic device and for displaying visual customer interfaces, such as application menu interfaces, application icon interfaces, and the like. The components of the electronic device communicate with each other over a system bus.
Through the above description of the embodiments, it is obvious for those skilled in the art that the method in the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can be implemented by hardware, but the former is a better embodiment in many cases. Based on such understanding, the technical solutions of the present invention essentially or contributing to the prior art may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (e.g., a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present invention.
That is, according to an embodiment of the present invention, there is also provided a computer-readable storage medium having stored thereon a program for executing the method according to an embodiment of the present invention, which when executed by a processor, performs the steps of the method.
From the foregoing, it will be appreciated that specific embodiments of the invention have been described herein for purposes of illustration, but that various modifications may be made without deviating from the scope of the invention. Those skilled in the art will appreciate that the operations and routines depicted in the flowchart steps or described herein may be varied in many ways. More specifically, the order of the steps may be rearranged, the steps may be performed in parallel, the steps may be omitted, other steps may be included, various combinations of routines may be made, or omitted. Accordingly, the invention is not to be restricted except in light of the attached claims.
List of references
[1]Sushkov O O,Sammut C.Local image feature matching for objectrecognition[C]//International Conference on Control Automation Robotics&Vision.IEEE,2011:1598-1604.
[2]Lowe D G.Distinctive Image Features from Scale-Invariant Keypoints[J].International Journal of Computer Vision,2004,60(2):91-110.
[3] Li thought, recognition of the image of the circuit board residue, Beijing Master paper of post and telecommunications university 2017.28-43.
[4]Harris C.A combined corner and edge detector[J].Proc Alvey VisionConf,1988,1988(3):147-151.
[5]Simon J.D.Prince.Computer Vision:Models,Learning,and Inference[M].Cambridge University Press,2012.
[6]Wang Z,Bovik A C,Sheikh H R,et al.Image quality assessment:fromerror visibility to structural similarity.[J].IEEE Trans Image Process,2004,13(4):600-612。
Claims (10)
1. A circuit board fragment image identification method based on feature points and structural relations thereof comprises the following steps:
step 1, extracting a fragment image I corresponding to a circuit board fragment to be identified0The characteristic points of (1);
step 2, calculating SIFT description of the feature points to form a feature point set S0；
Step 3, respectively calculating the feature point set S0With the SIFT description of each feature point in each sample image IiCharacteristic point set S ofiEach of (1) toFirst Euclidean distance of SIFT description of individual feature points, wherein the sample image IiCorresponding to a known circuit board, and the sample image IiAnd the characteristic points and the characteristic descriptions thereof are stored in a sample library;
step 4, sorting the first Euclidean distances from small to large, and selecting a feature point set SiThe first Euclidean distance of the plurality of characteristic points in the sequence of S is taken as0A candidate matching point of the feature point of (1);
step 5, respectively calculating each candidate matching point and S0A second Euclidean distance between the shape context descriptions of the feature points, and candidate matching points with the second Euclidean distance not greater than a predetermined threshold are reserved as S0The set of matching points of the feature points;
step 6, establishing the fragment image I0And in the sample image IiSearching a second star structure matched with the first star structure, wherein the second star structure is established based on the matching point set;
step 7, obtaining each sample image I according to the similarity of the second star structure and the first star structureiAnd a broken picture image I0The degree of matching of (2).
2. The method for recognizing the image of the circuit board fragment based on the feature points and the structural relationship thereof as claimed in claim 1, wherein the feature points in step 1 are Harris feature points.
3. The method for recognizing the image of the circuit board fragment based on the feature points and the structural relationship thereof as claimed in claim 1, wherein in step 6, the fragment image I is established as follows0The first star structure of the feature points of (1):
step 6-1, obtaining the fragment image I0Current feature point p ofkLet k equal to 1;
step 6-2, if the current characteristic point pkIs not in the residual image I0Does not belong to the determined image edge regionThe effective star structure of (2) is used as the center point of the star structure, otherwise, the next characteristic point p is usedk+1As the current feature point, repeating the step;
step 6-3, feature point set S0And the current feature point p inkIs not less than the star-shaped structure threshold TdIs defined as pkEnd node p of central star structuree；
Step 6-4, if with pkSaid end node p of the central star-shaped structureeNot less than a predetermined number, will be given by pkThe central star structure is used as an effective star structure, and the current central point p is taken askIs marked asEnd node peIs marked as
Step 6-5, if the next characteristic point p does not existk+1Then record all valid star structures and let the current center point pkIs marked asEnd node peIs marked asWherein M is 1, …, M is the number of effective star structures, if the next characteristic point p existsk+1Then the next feature point p is setk+1And returning to the step 6-2 as the current characteristic point.
4. The method for recognizing the image of the circuit board fragment based on the feature points and the structural relationship thereof as claimed in claim 3, wherein in the step 6, the sample image I is processed by the following stepsiSearching for a second star structure matched with the first star structure:
step 6-6, in sample image IiIs centralized and traversedThe matching point set of (2) takes each feature point of the matching point set, which is not at the edge, as the central matching point M of the star structurec；
6-7, in the sample image IiUpper center matching point (M)c) For starting point, a line segment is establishedTaking a rectangular neighborhood by taking the terminal point of the line segment as the center;
6-8, traversing all the feature points M in the rectangular neighborhoodeIf the following conditions are satisfied, the point is an end matching point:
a)Meandthe euclidean distance described by SIFT of (a) is not more than 0.6;
b) line segmentAnd line segment McMeThe included angle is less than 10 degrees, and the side length ratio of the short side to the long side is more than 0.97;
step 6-9, if the fragment image I0A plurality of end points of a certain star structure find matching points in the sample image, and then a group of star structure matching candidate pairs are formed.
5. The method for recognizing the image of the circuit board fragment based on the feature points and the structural relationship thereof as claimed in claim 4, wherein the steps 6-9 further comprise:
and taking the star structure matching candidate pair meeting the following conditions as a final star structure matching pair:
1) the number of the tail end matching points is not less than 3;
2) the angle difference between any two groups of edges of the two matched star structures is less than 12 degrees, and the adjacent relations of the characteristic points are the same;
3) any two groups of end points of the two matching star structures are connected, and the distance ratio of the short side to the long side is more than 0.87;
4) the structural similarity SSIM of two circumscribed rectangular image regions containing two star-shaped structures is greater than 0.3,
wherein SSIM (X, Y) ═ l (X, Y) c (X, Y) s (X, Y) (1)
Wherein,
wherein X, Y represents the pixel value arrays, μ, of the image sample image and the residue image, respectivelyXAnd muYRepresenting the mean, σ, of the imageXAnd σYRepresenting the variance, σ, of the imageXYRepresenting the covariance of images X and Y, C1, C2, C3 are preset constants.
6. The method for recognizing the image of the circuit board residue based on the feature points and the structural relationship thereof as claimed in claim 5, wherein in step 7, each sample image I is calculated as followsiAnd a broken picture image I0The matching degree S:
wherein i is the serial number of the star-shaped structure matching pair, eiIn order to match the number of sides of the star structure,for matching the square of SIFT description distance between feature points, SSIMmax is the star structure of the residue image and the sample image IiOf the matching star structure of (a).
7. The method for identifying the image of the circuit board fragment based on the feature points and the structural relationship thereof as claimed in claim 1, wherein the sample library is established as follows:
step 0-1, extracting and calculating each sample image IiForming SIFT description and shape context description of the feature points;
step 0-2, storing the characteristic description into a sample library, wherein each sample image IiSIFT descriptions, each with corresponding feature points, and shape context descriptions.
8. The method for recognizing the image of the circuit board fragment based on the feature points and the structural relationship thereof as claimed in claim 1, wherein during the calculation of the shape context description, a rectangular coordinate system is used to replace a logarithmic polar coordinate system, and the spatial distribution of each feature point within a certain window size centered on each feature point is counted to describe the structural relationship between the adjacent feature points.
9. The method for identifying the circuit board fragment image based on the feature points and the structural relationship thereof as claimed in claim 1, further comprising a preprocessing step before the step 1, wherein for the fragment image which can not extract the stable feature points, the region of the component object is extracted and abstracted into the feature points, the feature points are replaced by the features such as the related colors, and then the steps 2 to 7 are carried out.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a program for executing the method according to any one of claims 1 to 9, which program, when executed by a processor, performs the steps of the method.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810710880X | 2018-07-03 | ||
| CN201810710880.XA CN108961240A (en) | 2018-07-03 | 2018-07-03 | Destructor circuit board relic recognition methods |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110148133A true CN110148133A (en) | 2019-08-20 |
| CN110148133B CN110148133B (en) | 2021-04-06 |
Family
ID=64484702
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810710880.XA Withdrawn CN108961240A (en) | 2018-07-03 | 2018-07-03 | Destructor circuit board relic recognition methods |
| CN201910534181.9A Active CN110148133B (en) | 2018-07-03 | 2019-06-20 | Circuit board fragment image identification method based on feature points and structural relationship thereof |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810710880.XA Withdrawn CN108961240A (en) | 2018-07-03 | 2018-07-03 | Destructor circuit board relic recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN108961240A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111179323A (en) * | 2019-12-30 | 2020-05-19 | 上海研境医疗科技有限公司 | Medical image feature point matching method, device, equipment and storage medium |
| CN113160284A (en) * | 2021-03-09 | 2021-07-23 | 大连海事大学 | Guiding space consistent photovoltaic image registration method based on local similar structure constraint |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0438574A (en) * | 1990-06-05 | 1992-02-07 | Kowa Co | Pattern checking device |
| CN102289681A (en) * | 2011-08-05 | 2011-12-21 | 上海邮政科学研究院 | Method for matching envelope images |
| CN103778435A (en) * | 2014-01-16 | 2014-05-07 | 大连理工大学 | Pedestrian fast detection method based on videos |
| CN104102913A (en) * | 2014-07-15 | 2014-10-15 | 无锡优辰电子信息科技有限公司 | Wrist vein certification system |
| CN104281679A (en) * | 2014-09-30 | 2015-01-14 | 东软集团股份有限公司 | Goods classification method and goods classification device both based on image features |
| CN104463178A (en) * | 2014-12-29 | 2015-03-25 | 广州视源电子科技股份有限公司 | Electronic component identification method and system |
| CN104915949A (en) * | 2015-04-08 | 2015-09-16 | 华中科技大学 | Image matching algorithm of bonding point characteristic and line characteristic |
| CN105117054A (en) * | 2015-08-12 | 2015-12-02 | 珠海优特电力科技股份有限公司 | Handwriting input recognition method and system |
| CN105469402A (en) * | 2015-11-24 | 2016-04-06 | 大连楼兰科技股份有限公司 | Automobile part identification method based on space shape context characteristic |
| CN106373161A (en) * | 2016-09-21 | 2017-02-01 | 哈尔滨工业大学 | SIFT (scale-invariant feature transform) feature point-based positioning method |
| CN106780440A (en) * | 2016-11-29 | 2017-05-31 | 北京邮电大学 | Destruction circuit plate relic image automatic comparison recognition methods |
| CN107330928A (en) * | 2017-06-09 | 2017-11-07 | 北京理工大学 | Based on the Image Feature Matching method for improving Shape context |
-
2018
- 2018-07-03 CN CN201810710880.XA patent/CN108961240A/en not_active Withdrawn
-
2019
- 2019-06-20 CN CN201910534181.9A patent/CN110148133B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0438574A (en) * | 1990-06-05 | 1992-02-07 | Kowa Co | Pattern checking device |
| CN102289681A (en) * | 2011-08-05 | 2011-12-21 | 上海邮政科学研究院 | Method for matching envelope images |
| CN103778435A (en) * | 2014-01-16 | 2014-05-07 | 大连理工大学 | Pedestrian fast detection method based on videos |
| CN104102913A (en) * | 2014-07-15 | 2014-10-15 | 无锡优辰电子信息科技有限公司 | Wrist vein certification system |
| CN104281679A (en) * | 2014-09-30 | 2015-01-14 | 东软集团股份有限公司 | Goods classification method and goods classification device both based on image features |
| CN104463178A (en) * | 2014-12-29 | 2015-03-25 | 广州视源电子科技股份有限公司 | Electronic component identification method and system |
| CN104915949A (en) * | 2015-04-08 | 2015-09-16 | 华中科技大学 | Image matching algorithm of bonding point characteristic and line characteristic |
| CN105117054A (en) * | 2015-08-12 | 2015-12-02 | 珠海优特电力科技股份有限公司 | Handwriting input recognition method and system |
| CN105469402A (en) * | 2015-11-24 | 2016-04-06 | 大连楼兰科技股份有限公司 | Automobile part identification method based on space shape context characteristic |
| CN106373161A (en) * | 2016-09-21 | 2017-02-01 | 哈尔滨工业大学 | SIFT (scale-invariant feature transform) feature point-based positioning method |
| CN106780440A (en) * | 2016-11-29 | 2017-05-31 | 北京邮电大学 | Destruction circuit plate relic image automatic comparison recognition methods |
| CN107330928A (en) * | 2017-06-09 | 2017-11-07 | 北京理工大学 | Based on the Image Feature Matching method for improving Shape context |
Non-Patent Citations (3)
| Title |
|---|
| WENJUN LU 等: "Image Quality Assessment Based on SIFT and SSIM", 《CHINESE CONFERENCE ON IMAGE AND GRAPHICS TECHNOLOGIES》 * |
| 庄大燕: "人体手背静脉识别技术研究", 《中国优秀博硕士学位论文全文数据库 (硕士) 信息科技辑》 * |
| 李俊山 等: "异步多特征红外与可见光图像匹配算法", 《微电子学与计算机》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111179323A (en) * | 2019-12-30 | 2020-05-19 | 上海研境医疗科技有限公司 | Medical image feature point matching method, device, equipment and storage medium |
| CN113160284A (en) * | 2021-03-09 | 2021-07-23 | 大连海事大学 | Guiding space consistent photovoltaic image registration method based on local similar structure constraint |
| CN113160284B (en) * | 2021-03-09 | 2024-04-30 | 大连海事大学 | Guidance space-consistent photovoltaic image registration method based on local similar structure constraint |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108961240A (en) | 2018-12-07 |
| CN110148133B (en) | 2021-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107679448B (en) | Eyeball action-analysing method, device and storage medium | |
| Konstantinidis et al. | Building detection using enhanced HOG–LBP features and region refinement processes | |
| Terrillon et al. | Comparative performance of different chrominance spaces for color segmentation and detection of human faces in complex scene images | |
| US9367766B2 (en) | Text line detection in images | |
| Terrillon et al. | Automatic detection of human faces in natural scene images by use of a skin color model and of invariant moments | |
| CN110765954A (en) | Vehicle weight recognition method, equipment and storage device | |
| CN109740606B (en) | Image identification method and device | |
| CN106780440A (en) | Destruction circuit plate relic image automatic comparison recognition methods | |
| JP2011022991A (en) | Image recognition method and apparatus | |
| CN111860309A (en) | Face recognition method and system | |
| CN110148133B (en) | Circuit board fragment image identification method based on feature points and structural relationship thereof | |
| CN110427912A (en) | A kind of method for detecting human face and its relevant apparatus based on deep learning | |
| Lee | Component-based face detection and verification | |
| Kobchaisawat et al. | Thai text localization in natural scene images using convolutional neural network | |
| JP5791751B2 (en) | Image recognition method and image recognition apparatus | |
| CN107368830A (en) | Method for text detection and device and text recognition system | |
| Fritz et al. | Object recognition using local information content | |
| Güzel | Performance evaluation for feature extractors on street view images | |
| de Santos-Sierra et al. | Low computational cost multilayer graph-based segmentation algorithms for hand recognition on mobile phones | |
| Wei et al. | Omni-face detection for video/image content description | |
| Soni et al. | Extracting text regions from scene images using Weighted Median Filter and MSER | |
| Wei et al. | Detection of side-view faces in color images | |
| Hossain et al. | A novel approach to extract region from facial expression based on mutation | |
| Saputra et al. | Variance-Based Geometric Feature Selection for Face Recognition System | |
| Wiwatcharakoses et al. | MSER based text localization for multi-language using double-threshold scheme |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |